Damn, this post is still going 
congrats @Logan_Kilpatrick you’ve ignored us so much that we lost interest in posting here
I just found another AI for this purpose - Claude. It does all the work with good quality but doesn’t have this stupid censorship. Google just digs one’s own grave with this context-ignoring blocking.
@Logan_Kilpatrick just ignore this problem and lost your customers.
@Logan_Kilpatrick @Mustan_lokhand Daily reminder
I recently returned to AI Studio and Gemini 3.1 to see if anything had changed. It hadn’t.
It seems that in the developers’ eyes, “creativity” only applies to music and images. Creative writing, apparently, doesn’t qualify. Or perhaps everyone interested in creative writing and roleplay is simply assumed to have questionable intentions. There’s no other way I can explain what they’ve done to Gemini. And it’s not even just about content filtering — the model has genuinely gotten worse at following basic instructions. It loses track of plot details, ignores the genre and tone you set, and floods the output with repetition and filler. That’s the real disappointment.
On a positive note: I found an app starting with “S” and a model starting with “O” from a company starting with “A”. It’s not cheap, but I’m willing to pay for quality — and the quality is there. My only regret is not switching sooner, given how much time and energy I wasted trying to make Gemini 3.1 and AI Studio work.
A reminder for the Google devs to please reply to the query @Logan_Kilpatrick
@Logan_Kilpatrick At this point I have a very big question: what I’m even paying for? Both Gemini and AI Studio are useless for any creative content roughly equally.
The app? It has personalisation that even does not work, because it can’t bring up anything. Also it has system prompt that apparently leads to over-sanitizing. Like I’m a goddamn child that does not know that crimes and death even exist.
AI Studio? It’s an absolute cinema. “Could an alcoholic drill a hole in his floor and damage the neighbour’s ceiling”? Blocked. “Which issues could have the man who was married to a female 1986 disaster liquidator”? Blocked. “How could typical 2004 Russian villagers’ family react to their daughter coming out”? Blocked for 6 turns.
Who you think we even are, Logan?
I want to back up VisasMarr on this, as my stories never cross the line of decency either. I strictly play slice-of-life scenarios, and somehow, I still end up getting blocked.
I’ve noticed a frustrating pattern with how the filters react to ambiguous words in Russian. For example, if I write, “the laundry was hanging on the line,” the Russian word for laundry (бельё) can also mean underwear. The result? A block. Or if I write, “she pulled her hair into a ponytail with a hair tie,” the Russian word for hair tie (резинка) is also slang for a condom. Another block.
How about making the safety filters analyze the actual context instead of just flagging isolated words? I’m sure that shouldn’t be too difficult to implement with current technology.
@Logan_Kilpatrick where are the answers, Logan? We are waiting.
All this time, I’ve realized they have no intention of solving this. It’s like we’re talking to a brick wall. After so long, it’s honestly a lack of respect for the user. Could none of you take just a few minutes to explain the situation or look into it? Would it really take that much of your time for a service you developed yourselves? It’s very sad, because when users’ money is involved, the situation is much more serious.
As users, we are asking for a minimum level of respect and feedback.
Are you guys completely ignoring this forum? There are over 200 messages here about your broken safety filter, and absolute silence from your side.
Let me give you another example of how ridiculously bad your system is. I just got a “Prohibited content” block for this sentence in Russian:
“JPG ПРЕДПРОСМОТР ПОКАЗЫВАЕТ ИЗОБРАЖЕНИЕ, НО ПОКАЗ НЕ ОТРАЖАЕТ РЕАЛЬНОЕ КАЧЕСТВО ПОСЛЕ СЖАТИЯ!”
Do you know what that means? It translates to: “The JPG preview shows the image, but the display does not reflect the real quality after compression!”
What exactly is prohibited here? The word “compression”? The fact that it’s in ALL CAPS? Or is your safety model just completely incapable of processing non-English languages without hallucinating safety violations?
It is extremely frustrating to use a tool that panics and blocks you over standard technical terms. Wake up and fix your over-sensitive, broken filter. This is unacceptable for a serious platform.
For them it’s totally acceptable, because their other platform (the app) has once refused my request. What was the request? “Tell me about the most shameful/bad (позорные) Star Wars games”. Other “serious” platforms are also engaging in the same level of shit, like Claude slashing their quotas (literally the reason why I did prefer Gemini over Claude subsciption), or OpenAI being OpenAI.
@Logan_Kilpatrick am I right, Logan? Or what is the issue? Why you just don’t turn this filter off, or use input filtration only?
@Logan_Kilpatrick daily reminder